How many launches showed up, how many passed the filters, and what type of setup appeared.
Student notebook for on-chain markets
Learning Solana launches in public.
I am building a small system that watches new token launches, filters obvious risk, and tracks what happens after. The goal is not to call coins. The goal is to learn from the data.
Today
signals
mistakes
changes
Latest result
Day 9: First red day, useful regime data.
SOL dropped about 9%, and the same mean-reversion bot went from a strong green day to basically flat. This is the risk-off regime test happening live.
Trades
11
Win rate
72.7%
Net P&L
-0.0004 SOL
EV / trade
-0.15%
Main lesson
- Real on-chain trades only: 8W / 3L, 8 TP / 3 SL
- Average win: +3.36%; average loss: -9.48%
- Profit factor: 0.9; median hold: 24 minutes
- Three larger stops erased eight small wins in a downtrend
Previous results
Day 8: First strong real-money session. 14 trades 92.9% win +0.0113 SOL
The bot ran fully automated for 18 hours with real SOL. Entries, sizing, swaps, TP, SL, and exits all ran on-chain while I slept.
+3.45% EV/trade
13 TP / 1 SL
Profit factor 7.3
Day 7: Same entries, smaller exit fits better. 22 live trades 50% win -0.00054 SOL
The live TP 5% baseline was basically breakeven. A paper-sim on the same entry signals showed TP 2.5% captured more of the fast pump before the fade.
TP 5% live
TP 2.5% sim: 79% win
Exit fit improved
Day 6: First real live session, red but useful. 9 live trades 29% win -0.00589 SOL
The first real SOL v3 session lost money, but validated the infrastructure: swaps landed on-chain, and both TP and SL fired automatically.
7 closed / 2 open
-4.2% EV/trade
Execution worked
Day 5: Small paper test, same v3 edge. 5 trades 80% win +3.17% EV
Last night was a longs-only paper test on tier 3-4 Solana memes. The sample was tiny, but the behavior matched the broader v3 validation.
4 TP / 1 SL
+0.00317 SOL
Paper mode
Day 4: New lane, paper test turned green. 45 trades 66.7% win +0.0327 SOL
I rotated from newborn pump.fun launches into tier 3-4 mid-cap Solana memes. The first v3 paper session was positive, but still needed live execution-cost validation.
+1.32% EV/trade
p=1.8%
Paper mode
Day 3: Near-50% win rate, still negative EV. 10.8h 42 trades -0.2949 SOL
Execution captured most alerts, but average losses were much larger than average wins, so the larger-size simulation stayed red too.
48.8% win rate
-13.3% EV/trade
92 AI evaluations
Day 2: Execution cleaner, still testing small. 8.8h 32 trades -0.1539 SOL
Direct PumpSwap execution improved sell reliability. The live test stayed red at tiny size, but execution got cleaner.
Fees dominated
Execution improved
Still testing small
Day 1: Bad result, useful data. 10.5h 48 trades -0.249 SOL
First real session ran for 10.5 hours. The detector found movement, but execution friction ate the edge.
44 closed
Fixes shipped
Execution friction
Early
learning phase
Daily
results posted
Public
mistakes included
No calls
notes only
Daily note format
Simple enough to post every day.
The best outcome, the worst outcome, and the part of the move that looked obvious only after the fact.
One small adjustment to the system, the process, or the way I read the data tomorrow.
The system
A small detector, a daily notebook, and a lot of testing.
Detect
Watch new Solana token launches and keep the useful ones in view.
Filter
Remove obvious risk before pretending a chart means anything.
Track
Follow post-launch behavior and compare the signal to the outcome.
Learn
Post the results, keep the mistakes visible, and improve the next test.
Principles
Building in public without pretending to know everything.
- No live buy calls
- No paid group
- No guaranteed results
- Feedback welcome